1HBM sold out into next year, and the choke points are in the equipment, not the chips
High-bandwidth memory prices roughly doubled in a quarter, and the makers — Micron, SK Hynix, Samsung — are sold out into next year. Brian, on Business with Brian, takes that as the setup for the wrong trade: buying the makers means paying for a shortage that is already in the price, while the parts of the chain that cannot be substituted sit one layer down. He maps five of them, each a choke point — interface (), tools (, ), inspection (, ), test (, ), and a foundation layer of bonders and materials (, , ).
The fundamentals he cites are what make it more than a diagram. Rambus's operating margin went from 9% to 37% as the revenue mix turned royalty-driven. KLA's revenue is up 76% and operating income up 110% over four years, from a near-monopoly position in inspection. Advantest's sales are up 45% with operating profit up 120%. Besi's new orders are up 104% year on year, driven by hybrid bonding for HBM4 parts that ship in 2027. Entegris draws roughly 30% of its revenue from the most advanced 5% of wafers made — the narrowest slice of the industry, and the one that grows.
He is explicit about where it is fragile. is the speculative leg: revenue down 57% over four years, roughly break-even today, and the whole case is a 2027 forward story. gets the smallest slice of his model book — $8 of $100 — because an open antitrust inquiry sizes it down. is named as the obvious way to own memory and then explicitly passed over. Two of the twelve names trade as over-the-counter ADRs (, ), one of them flagged as thin.
The tape was pulling the same way without the layer distinction. Marvell was up 9% pre-market on its S&P 500 inclusion, Intel up 6.5% and AMD up 4.7%, with Qualcomm bid on a ByteDance deal and Broadcom carried along on the same semiconductor breadth. Super Micro went the other way, down 4.7%, so the bid was broad rather than indiscriminate. What none of this dates is the memory argument itself: it is framed as a five-year build in its first innings, which means no print this week marks it either way. This is also one video's argument on a day when five of the six channels checked published nothing, so it is one well-evidenced voice rather than corroboration — and that video carried a disclosed paid promotion for a junior gold miner, separate from the memory work.
2A preliminary Iran deal took the war premium out of oil before a barrel moved
The most-repeated item in the day's news by a wide margin was a preliminary US–Iran peace deal and the prospect of the Strait of Hormuz reopening — a dozen appearances or more across Yahoo Finance, Finnhub and Google News, against four for the next-biggest theme. Crude fell to a three-month low on it, and equities took it as risk-on.
The gap is between the paper and the barrels. The coverage describes the deal as fragile and preliminary, with signing set for Friday, and both Hormuz mine-clearance and the timing of actual flows unresolved behind it. The price has moved on an outcome whose logistics have not started. That asymmetry runs one way: the low is already banked, so what reprices it is a slip, not a success.
The risk-on read is not clean either. Gold rose 2.6% in the same session with the dollar at a 10-day low — a pairing the day's sources file under a dovish reassessment of the rate path rather than under geopolitical relief. Both readings fit the tape. What separates them is whether gold keeps its bid once the headline is spent; if it does, the move was about rates, not about Iran.
For energy the read stayed at the sector level rather than resolving into a name: a three-month low in crude on a de-escalation headline is a headwind to anything carrying oil exposure, and the day's material stopped there.
3Space's biggest debut came with double-digit falls in its neighbours
SpaceX debuted up 19%. In the same session Firefly fell 19% and AST SpaceMobile fell 15.5%, and the day's coverage frames that pairing as rotation inside the sector rather than new money into it. The headline number for space was up; its two listed incumbents were not.
The consequence is about position rather than direction. If a sector reprices internally when one name arrives, then owning the sector is not the same trade as owning the right part of it. The evidence is one session of pre-market prints from a single source, so it sizes an observation and not a trend — what would settle it is whether and recover once the listing settles.
What the sources recommended
Thirteen buys and no single-name sells — and twelve of the thirteen are one modelled allocation from one video, not twelve independent calls.
The buys. Brian's memory-stack book is a $100 model portfolio rather than a list: $25 to tools (, ), $25 to inspection (, ), $22 to test (, ), $20 to the foundation layer (, , ), and $8 to , sized down for the open antitrust inquiry. Read the twelve as one bet on memory-supply choke points held over months to years, with carrying his own speculative label. He adds two funds with the caveat attached in the same breath: holds the makers only and blends makers and suppliers, and he says outright that the pure supplier basket he is arguing for is packaged by neither — so both are compromises on his own thesis.
The thirteenth is unrelated to any of it. (Oracle) picked up an analyst upgrade to Buy from Zacks, carried by Yahoo Finance. That is a rating change rather than a thesis, and nothing else in the day's material develops it.
The three-month low in crude gets its test at Friday's signing
The Iran deal is the only dated catalyst the day produced. Signing is set for Friday, with mine-clearance in Hormuz and the timing of flows still unresolved behind it — so the paper and the barrels are on different clocks, and crude's three-month low is priced off the faster one. A slip in the signing, or a slower clearance than assumed, is what puts that low back in play.
The memory argument has no such date. Its dated part sits in 2027, when the HBM4 hybrid-bonding orders Besi is booking today turn into shipments, which is why nothing between now and then confirms or breaks it.
